Llama-3.3-70B

Measured 2026-08-25

* Conditional

Fails restraint — fires a tool with a made-up ID instead of asking for the missing one.

Score

Task done

64.21%

43–84% 95% CI bootstrap over items · n=95 · measured 2026-08-25

Refusal quality

20.0%

names the gap 33% · offers a route 13% · 10 probes x 3 iterations · 19 silent turns scored zero

Fabrication

0 of 10 probes (30 probe runs) · ≤31%

Dead-air

74.3%

260 of 350 turns

Tool silence

100.0%

all of 19 items (260 tool-call turns) · ≥82%

Stalled

26.7%

24 of 90 runs

TTFT p50

698ms

698–860 p50–p90

Cost / 1M tok

$1.04

Share this result

Download card (PNG) Post on X

Embed the live badge

Markdown

[![Speko llm rank](https://benchmarks.speko.ai/badge/llm/together-llama-3-3-70b.svg)](https://benchmarks.speko.ai/llm/together-llama-3-3-70b)

HTML

<a href="https://benchmarks.speko.ai/llm/together-llama-3-3-70b"><img src="https://benchmarks.speko.ai/badge/llm/together-llama-3-3-70b.svg" alt="Speko llm rank"></a>

URL

https://benchmarks.speko.ai/llm/together-llama-3-3-70b